You are on page 1of 34

B of Governors of the Federal Reserve System

InternationalFinance DiscussionPapers

Number 567

October 1996

FINANCIALINNOVATIONAND THE SPEED OF ADJUSTMENTOF MONEY DEMAND:
EVIDENCE FROM BOLIVIA,ISRAEL, AND VENEZUELA

Martina Copelman

NOTE: InternationalFinance DiscussionPapers are preliminarymaterialscirculatedto stimulate
discussionand critical comment. Referencesin publicationsto InternationalFinance Discussion
Papers (other than an acknowledgmentthat the writer has had access to unpublishedmaterial) should
be cleared with the author or authors.
BSTRACT

Traditionalstudiesof money demand for both developedand less developedcountrieshave

shown that there are periods of “missingmoney,” that is, there is consistentoverpredictionof real

balances. This paper uses cointegrationtechniquesto study the effects of financialinnovationon the

demand for real balances in Bolivia,Israel, and Venezuela. The results show that financialinnovation

can accountfor the instabilityof money demand observed in these countries. In particular, I find that

the long run demand for real balancesshified down. In addition, I show that the speed at which

people adjust their demand for money when out of equilibriumincreasesfollowingfinancial

imovation.
FinancialImovation And The Speed of Adjustmentof Money Demand:
EvidenceFrom Bolivia,Israel, And Venezuela.

Martina Copelmani

ION
1.INT’RODUCT
Empiricalstudiesof the demandfor moneyhave flourishedin recent years. The reason for this

largevolumeof literaturecan be explainedby the importanceof moneydemandin monetarypolicy. The

incomeand interestelasticitiesof moneydemandare at the core of the mostbasicmacroeconomicmodels

such as the IS-LM; where the effectivenessof monetarypolicy (that is, its abilityto affect the real side

of the economy)depends on the elasticityof money demand. It is very importanttherefore, to find

consistentestimatesof these elasticities.

Following the seminal papers by Goldfeld (1973,1976) a large number of papers for both

developedand less developedcountrieshave shownthat there havebeen periodsof “missingmoney”that

is, there is consistentoverpredictionof real balances. This has led to the conclusionthat the money

demand functionis basicallyunstable.

Two principalexplanationshavebeen used to accountfor this instabilityof moneydemand. One

is financialimovation and the other is currencysubstitution. Throughoutthe 1980’schangesin financial

markets have been very widespread, and have been particularly important in Bolivia, Israel, and

Venezuela. All three of these countriessufferedhigh inflationsduring the 1980’s;two of them, Bolivia

and Israel, were actually hyperinflations. In the mid to late 1980’s Bolivia, Israel, and Venezuela

‘ The author is staff economistin the Divisionof InternationalFinance, Board of Governorsof the
Federal Reserve System. The views expressedin this paper are solely the responsibilityof the author
and shouldnotbe interpretedas reflectingthoseof the Boardof Governorsof the FederalReserveSystem
or other Members of its staff. I am grateful to Rudiger Dornbusch, Stanley Fischer, David Hendry,
Robert Solow, Alejandro Werner, and the participants of the M.I.T. InternationalBreakfast and the
MoneyLunch as well as the participantsof the seminarsat the Federal ReserveBoard and the XII Latin
American EconometricSocietyMeetingsfor helpfil commentsand suggestions.
embarked on stabilizationprograms designedto halt inflation.

The purpose of this study is to show that financial innovation leads to a faster speed of

adjustmentof moneydemandto its determinants,as well as to the instabilityof the observeddemand for

real balances. This is an aspect of the money demand literature which has hardly been explored, in

particular in terms of the new cointegrationmethods. For Israel, Melnick (1991) shows that ignoring

financialinnovationis theprincipal cause for the lack of stabilityin the money demand equation. I will

showthatthe long run demandfor real balancesnot onlyshifieddown, but that in the short run the effect

of financial imovation has-been to increase the speed with which people adjust their actual money

holdingsto their desired money holdings.

Engle and Granger (1987)have shownthat using the traditionalmethodof partial adjustmentto

estimatethe demand for money may lead to misspecificationand incorrectconclusions. It is important

to take accountof the fact that the variablesused in the estimationare non stationary, somethingthat is

usuallyignored. Ignoringthe stochasticpropertiesof real moneybalances,the transactionsvariable, and

the opportunitycost variablecan lead to OLS regressionresults which are statisticallyinvalid. For this

reason, I willuse whathas become the standardcointegrationapproachto estimatethe demandfor money

for Bolivia, Israel, and Venezuela.

The remainder of the paper is structured as follows. Section 11presents the model which

provides the theoretical foundationsfor the empiricalsection. I use the Miller-Orr (1966) model of

moneydemandand showthat a declinein the transactionscost (whichis a proxy for financialimovation),

leadsto a faster adjustmentof moneydemandto its determinants. Section111providesa brief description

of the events in Bolivia, Israel, and Venezuela. Section IV presents the cointegration methodology. In

SectionV I estimatethe long run demandfor moneyin allthree countriesusing the cointegrationmethods

of the previous section. In this section I also test for the stabilityof the demand for real balances and
,.-
allowfor both a one time shift and a changein the slope. In SectionVI the short run demandfor money

is estimatedas an error correction model and the change in the speed of adjustmentis demonstrated.

2
SectionVII concludes.

SECTIONII. THE MODEL

The long run money demand functionis representedin terms of the Miller-Orr (1966)model.

The model is a transactionsdemandfor money where the decisionmaker can hold two types of assets,

1) money which does not earn any interestand 2) a “bond”or other earning asset whichpays interestat

a rate i per dollar per day. In what followsI will use the inflationrate, Zt , as the relevantopportunity

cost, since in all the three countriesin the study interestrates were fixed for a large part of the sample,

The firm faces a stochasticcash flow that can be characterizedby a sequenceoft independent

Bernoullitrials per day. In each trial cash balancescan go up by x with probabilityp or can go down

by x with probabilityq= l-p; where x is the “stepsize.” Transfers of funds betweenthe two assets can

be made instantlyat costb. This is a cost which is incurredper transactionand is independentof the size

of the transaction. A reduction in b decreases the cost of transactionsand therefore, I will associateit

with financial imovation as in previous studies. Finally, the minimum level of money holdings is

normalizedto zero.

The time path of cash holdingswill evolveas in Figure 1, where h is the maximumholdingsof

cash that the individualwill ever have, and z is the levelto whichbalancesare restored after a transfer.

The optimaldesired level of cash holdingscan be derivedby minimizingthe total expecteddaily cost of

this Ss policy given the two parametersh and z.”

The solutionto this problem is an optimallevel of cash holdingsgiven by

h“ +=” .4 3b (1)
m,” = ~ [~ ~’~1’”
3

where

See Miller and Orr (1966)pages 420-423for the methodologyand derivationof equations(1) and
(2).

3
i

,’

h

.-

z

o

4
~ * = ~3 b X2 t1,,3 , ~ * =3Z* (2)
4n

and where ~“ denotesthe optimaldesired stock of real moneybalancesand Ztis the inflationrate which

is a proxy for the opportunitycost. Equation (2) gives the optimalupper bound as a function of the

optimalreturn point. With this equationfor the optimallevelof cash balancesin mind, we now proceed

to determinethe effects of financialinnovationon both the speed of adjustmentand the level of optimal

money holdings.

In this context,financialinnovationis proxied as a declinein the transactionscosts, b. A lower

transactionscost impliesa lower real cost of transformingfinancialassets into money. Holdingall else

constant,this impliesthat financialinnovationinducesa reductionin the desired quantityof real money

balances that are held. One can actually think of b as also including effects associated with the

introductionof more instrumentsat the time of the financialliberalizationwhich can be used to avoid

holdingmore cash balances.

Equation(2) showsthat a reductionin b will decrease h*and z*,narrowingthe band of inaction

of Figure 1; while equation (1) shows that m*also declines. The idea is that when transactionscosts

decline,an individualwill wait less timeto adjusttheir moneybalanceswhen they are out of equilibrium.

Therefore, as agentsadjustsooner than before, the speedof adjustmentof desired moneyholdingsin the

economyas a whole increases.

Next, we take a brief lookat the financialliberalizationepisodesthat tookplace in each country.

SECTIONI A LooK AT BOLIVIA.ISRAEL. AND VENEz~
11.

111.1BOLIVIA

During 1984-1985,Boliviaexperiencedthe highest inflationrate of Latin America and one of

5
the highest in world history. Using Cagan’s (1959)definitionof a hyperinflation,that is 50% a month

or higher, Sachs (1986)places Bolivia’shyperinflationas the seventhhighestof the twentiethcentury.

During approximatelythe year and a half which the hyperinflationlasted, the averagemonthlyinflation
.,
rate reached46%. In responseto this, the Boliviangovernmentintroduceda stabilizationprogramin late

Augustof 1985. The hyperinflationquicklysubsidedand over the past few years has remainedboth low

and fairly stable. This program was introducedin conjunctionwith an extensiveliberalizationof all

markets, and financialmarkets in particular. -

Morales (1988)describesthe vast financialliberalizationwhich was introducedin Bolivia. All

interest rate ceilingswere eliminatedand capitalmarket restrictionswere also eliminated. Bankswere

allowedto operatein internationaltrade and capitalaccounttransactionswithoutrestrictions. In addition,

depositorswere allowedto opendollarand dollar indexedaccountsand bankscouldmakeloansin dollars

or indexedto the dollar as well as any localcurrency loan they deemedvaluable. Thesedollar accounts

have lower reserve requirementsthan other accounts.

The effect of this financial liberalizationwas to expand the definition of M2, of which the

depositsin dollars were the principalcomponent. Short term dollar deposits and those indexedto the

dollar increaseddramatically,from less than $28 millionin Septemberof 1985to $270millionin March

of 1987. The large shift in people’sportfolioswas mostly due to the very high interestrates that were
.
observed duringthis time, not only in pesos, but in dollars as well. Interest rates reached 32% in

September1985and remainedhigh for sometime. Table 1showsthat dollardepositsreachedabout78%

of total commercialbank depositsin 1989,up from less than 30% prior to the deregulation. Even after

the hyperinflationended,the demandfor real balancesnever returnedto its previouslevel(see Figure2),

which is firther evidencefor the hypothesisthat financialimovation causedthe declinein real balances.

6
TABLE 1.BOLIVIA: Dollarized Deposits and Interest Rates

1986 1987 1988 1989
l. DollarizedDeposits 45.5 67.5 73.4 78.0
(%oftotalcommercial
bankdep.)
.
2. Interest Rates 14.3 15.6 15.5 14.7
annualnominalratefordollarized
dep.
source: Morales (1991)

BOLIVIA: INFLATIONAND REAL Ml

200 2500

150

-.

tv
I
N
t It
I
#l
I II
It
I
I
I /,
I
1;,
//
I

*
50 *

0 I I I
Ir 1 I
I:
80 81 82 83 84 ’85 “86 ’87 ’88 ’89 ’90 ’91

l— Inflation ----- ml I
.
FIGURE 2A

7
FIGURE 2B

BOLIVIA: REAL Ml TO GDP RATIO

0.08

0.07

0.06

0.04

0.03

0.02 i I I I I I I
I
1 1 I
I
I I I
I
1 I I
I
I I I
I
I I I
I
I I I
I
1 1 I
I
I I I
I
I i I
I
I I

80 81 82 83 84 85 86 87’ 88 89 90 91
TIME

111.2ISRAEL

In July of 1985, Israel introduceda stabilizationprogram which was extremely successfulin

endingthehyperinflation. Unlikethe Boliviancase, Israel did not immediatelyadopt all capitalmarket

reforms. The drop in real balances experienced by Israel occurred in the beginning of 1987 in
conjunctionwith the adoptionof several large and importantfinancialmarket reforms.11

Previous studiesof money demand in Israel ( see for exampleMelnick (1991) and Ben-Bassat

and Marom (1988)) have found a shifi in the demand for real balances after the 1976 financial

liberalization,however, most studieshave not lookedat the period after 1983. Figure 3 showsthat real

balancesin the late 1980’swere lower than in the early 1970’s,even though inflationwas down to the

same levels and GDP was higher.

During the last quarter of 1986, Israel began a process of financial liberalization. The

liberalizationhas taken the form of a unificationand a reductionof mandatory liquidityratios. Most

importantly,in April of 1987non financialfirms were given full freedom to issue bonds. In addition,

limitationson foreign-exchangelinkedcreditwere relaxed. In essence,the government’srole in financial

markets was severely reduced (Ben-Bassat1988). This Iedtoa narrowing of interest rate gaps in the

capitalmarket and a decrease intheaverage rateon governmentbonds. Since the deregulationand the

stabilizationprogram, new financialinstrumentshave been a big part of Israel’s money market, which

have led to the decline in desired money holdings.

llIn mid 1985, there was a change in the regulationsdesignedto actuallyincrease money demand,
however, it was not succesfulas can be seen in Figure 3 and documentedin Ben-Bassat(1988).

9
I

FIGURE 3A

ISRAEL: INFLATION AND REAL BALANCES

0.6 60
,
*I
II
***,4 ,
9- I

0.5 t
I
I I
I 50
t

8
●** \
.

0.4 40

0.3 30
I
f

I
I

*.’l
0.2 t
I
20
*
#
I
I
I

0.1 10

0.0 \ll[llllllllIiiIiiIll 1,11,,1,,,[,,11,,,11,, [,11][,,,,,,,,,,,,,,,, ,,,,,, I 1I
1“’1’f
0
70 72 74 76 78 80 82 84 86 88 90

—INFLATION ----- Ml

...

10
FIGURE 3B

ISRAEL: REAL Ml TO GDP RATIO

0.40

0.35

0.30

0.25
&
5 0.20
E
A
0.15

0.10

0.05

0.00 >111, ,1[, ,, [,,,111 ,1, ,,, ,1,,,,,,,,,,, ,,, ,,, ,,, , I I I
l’”
I I I
l’”
I I i
11”
I 1 1 1 I I I f

70 72 74 76 78 80 ’82 ’84 ’86 ’88’ ‘go
TIME

111.3VENEZUELA
In 1989Venezuelaexperienceda sharpdropin realmoneyl)alancesthatcannotbe accounted
for withthe traclitional
set of explanatoryvaria?)les.Figure4 shoWsrealmoneybalancesanciinflation

duringthisperiod. Thedropinmoneydernandcoincideswiththebeginningofafinancialliberalization
packageintroducedin1989alongwiththe stabilizationprogramwhichfloatedthe exchangerate and

removedall capitalrestrictionsandaccessto the foreignexchangemarket.

11
i

FIGURE 4A

VENEZUELA: INFLATION AND REAL Ml

35 100000

;
30 II
II
90000
II
It
●- .

25 80000

20 70000 ~
15 60000 ~

10 .’ 50000

5 40000

0 % I 1
1’ 1 1
1“’1’’’1’’’1’”1 “’1’ ’’1’ ’’1 ’’’1” ‘1’’ ’1 ’’’1” 30000
78 79 80 81 82 83 84 85 86 87 88 89 90 91

l— inflation ..--- Mll
FIGURE 4B

VENEZUELA: REAL Ml TO GDP RATIO

0.9

0.8

0.7

~
u 0.6
R
0.5

0.4

0.3 ? 1’ 1“’1’’’11’’1’”1’
I 1 I I
“11’’1’’’1’’’1’”1 “’l”f
78 79 80 81 82 83 84 85 86 87 88 89 90
TIME

Restrictionson nominal interest rates were lifted, new types of financialinstrumentssuchas

savingsand time depositaccounts,and new CentralBankbonds called “BonosZero Coupon”(or zero-

couponbonds)were introduced. The latteraccountedfor 55.7%, in 1990,0ftotalflows inthegross flow

of transactionsinthe stock market. In addition, the reserve requirementsfor demand deposits were

unified and the role of the Central Bank as a lender of last resort was firmly established.

These reforms led to nominalinterest rates of around 30% and to a strong decompositionof

portfoliosby agents in favor of highly liquid assets with very high returns, such as time deposits and

savings certificates. In 1990 alone, time deposits and savings accounts increased at a real rate of

13
approximately 119%. (See the Country Profile for Venezuela) This feature of the financial market

refo~s is not exclusiveto Venezuela. All three countriesintroduceda wide range of new instruments

with which people could save on money holdings. These new instrumentscreate irreversibilityeffects

on the demand for money (Piterman (1988)). This is one of the channels through which financial

imovation has permanenteffects on money demand.

Having looked at the financial imovation episodes in the three countries, I now turn to the

estimationof the long run and short run money-demandfor each country, In the next section, I briefly

discuss the cointegrationapproachto be used in the estimations.

SECTION IV,THE COINTEGRATION
N APPROACH

Traditional estimations of money demand have a partial adjustment specification which is

estimatedby OLS such as:

(3)

ConventionalOLS analysisassumesthat the error is not seriallycorrelatedand is not correlatedwith the

regressors. More importantly,it assumesthatall the regressorsare eitherdeterministicrandomvariables

or stationary. These assumptionsgive consistent OLS estimates and allow the use of t-statistics to

determinethe significanceof different coefficients.

- A problem arises, however, when the regressors are generatedby a nonstationaryprocess. In

this case, the OLS estimatesof equation(3) willnot be consistentand regression resultswillbe spurious.

This impliesthat the distributionfor the t and F statisticswill divergeas the samplesize increasesgiving

incorrectcriticalvalues (Phillips(1986)). In order to estimatethe long and short run demandfor money

correctly and be able to draw inferenceson the elasticities,we must use the cointegrationapproach.

The cointegrationapproach”deals with the case in which a linear combinationof nonstationary

variables is stationary. In general, a series X is integratedof order d if after differencingd times it is

14
stationary(X-I(d)). Throughoutthe paper I will refer to series which are I(l) or differencestationary

(d=l); that is after differencingoncethey are I(0). Let X,be a vector of independentvariablesincluding

a constant

Zt=y,-p fxt t=l,2,3. .. (4)

then, if ~ is the vector which will make z~stationary we say that X( and Y~are cointegrated with

cointegrationvector ~(which need not be unique ifXL is multivariate). This cointegratingrelation

(Y,=~’XJ is interpreted as the long run relationship between Y and X, and z, is known as the

“equilibriumerror”, since it measures the degree to which the system is out of equilibrium.

Sincethe error is stationary,the relationbetweenX and Y will return to the mean even though

each series individuallywillmovewithoutthattendency. Allowanceis made for the possibilityof serially

correlatedbut temporary divergencesfrom this relationship. If Xand Yarecointegrated, then OLS is

the appropriatemethod to estimatethe long run relationshipamong them, and the coefficientestimates

will be consistent(Stock 1987). However, the test statisticswill be meaningless. The long run money

demand finction can then be estimatedby OLS if it is a cointegratingrelationship.

The cointegrationapproachcan also be used to estimatethe short run demandfor money. The

representationtheoremby Engle and Granger (1987)indicatesthat series which are cointegratedcan be

representedby an error correctionmodel(ECM). This ECM representationcorrespondsto the short run

relationship among these variables consistent with the long run cointegratingequation. The ECM

representationis as follows:

where z~.lis the “equilibriumerror” from the cointegrationequation. Sinceall the variablesin the ECM

are I(0), OLS estimationprovides consistentestimatesand good test statistics. In equation (5) agents

15
marginallyadjust yt in responseto lagged changesin itself and the Xt variablesas well as disequilibria

from”the long run cointegratingequation(4). In this ECM framework, the matrix d can be interpreted

as a measureof the speedwith whichthe systemcorrectslastperiod’sequilibriumerror. For this reason,

it is usually called the adjustmentmatrix.

SECTION V: THE LONG RUN DEMAND FOR MONEY

The datafor all the countriesare quarterlyfromdifferentsources. For Bolivia,thedata are from

1980:1to 1991:4from UDAPE. The data for Israel are from the Bankof Israel from 1970:1to 1990:4.

The data for Venezuelaruns from 1978:1to 1991:4and come from various IFS issues.

V.1 Testing for Unit Roots.

Prior to estimating the long run demand for real balances, each of the variables must be

submittedto a unit root test since cointegrationmethodsrequire that all the variablesbe I(l). To test

the null hypothesisthat each variable is nonstationaryI ran both Dickey Fuller (DF) and augmented

DickeyFuller (ADF) tests on each of the followingseries: m is the log of real Ml (deflatedby the CPI),

y is the log of real GDP, and n is the log of the inflationrate.lll The general form of the test for any

variable X is:

The DF test omitsthe summationand both of the tests are run with and withouta trend. The numberof

lags, k, is four (4) which is commonfor quarterlydata. To rejectthe nullwe require p to be significantly

negative. Tables 2A to 2C show the results of the tests for the three countries.

“The data for Israel were provided by Rafi Melnick from the Bank of Israel, and for Boliviaby
UDAPE ( Unidad de Analisisde PoliticasEconomical)in La Paz.

1llForVenezuela,y is the log of real non oil GDP. All the series are seasonallyadjustedexceptthe
inflationrate.

16
TABLE 2A: BOLIVIA-UNIT ROOT TESTS 1980:1-1991:4

Variable DF(t) ADF(t) DF(nt) ADF(nt)
---------------------------------------------------------------------------------------
Ieveh
Y -3.24 -0.63 -3.25 -0.96
7K -2.80 -2.25 -2.59 -1.90
m -1.09 -0.88 -1.75 -1.53
lst. Difference
Ay --- --- -12.85 -2.50
An --- --- -10.55 -3.25
AM --- --- -7.71 -2.18
---------------------------------------------------------------------------------------
Critical Values5% -3.50 -3.50 -2.9 -2.9

TABLE 2B: ISRAEL-~IT ROOT TESTS 1970:1-1990:4

Variable DF(t) ADF(t) DF(nt) ADF(nt)
levels

Y -4.53 -4.66 -1.40 -1.97
n -2.68 -2.49 -2.75 -2.58
m -0.06 -1.51 -0.86 -1.74
lst. Difference
Ay --- --- -13.19 -3.37
An --- --- -12.33 -4.10
AM --- --- - 6.30 -2.58
Critical Values5% -3.50 -3.50 -2.9 -2.9
TABLE 2C: VENEZUELA-UNIT ROOT TESTS 1978:1-1991:4

Variable DF(t) ADF(t) DF(nt) ADF(nt)
---------------------------------------------------------------------------------------
levels
Y -2.86 -3.09 -0.76 -0.39
n -4.20 -2.54 -3.64 -1.96
m -1.61 -1.94 -1.00 -1.09
lst. Difference
Ay --- --- -11.83 -3.19
An --- --- -9.10 -5.13
AM --- --- -5.27 -3.02
---------------------------------------------------------------------------------------
Critical Values 5% -3.50 -3.50 -2.9 -2.9

The evidencefrom these tests shows that we cannotreject the null hypothesisof a unit root in
the levels of y, n, and m. There are a few exceptionsfor each countrybut by far the results using the
ADF test show that these variablesare I(l). For the first differences, all the tests reject the null at the
5% level. The exceptionsin each of the cases almost always involvesthe DF test which is the least
powerfulof the two tests. Under this criterion, and taking into accountthe low power of these tests, I
concludethat these series are nonstationary. Since the regressors of the money demand function are
generatedby an I(1) process, applyingOLS analysisto it will lead to misleadingresults. The correct way
to estimatethe money demand functionis to use tie cointegrationapproach.

--V.2Cointe@ation Tests
The evidenceabove shows that these variables are not stationary, therefore to run OLS on a
typicalpartial adjustmentspecificationwouldnot yield the appropriatecoefficientestimates. To estimate
a long run money demand for Bolivia,Israel, and Venezuelathe followingcointegrationequationwas
estimatedby OLS for each country.

m, =~0+~IY,‘p2nf ‘Pf p,>o,p2<o (’7)
..

To test for cointegration(i.e. to test for the existenceof a long run equilibriumrelationship),I

18
ran ADFandDF testson p~(the “equilibrium
error”)to see if it wasstationary.we 1 of Tables3A
to 3C shows the results. The evidencefor all three countries fails to reject the null hypothesisof RQ
cointegration(i.e. that p~is I(l)) for all the tests. This means that equation(7) is not a stable long run
money demand functionfor any of the three countriesstudied.
This result can be interpreted in terms of the model presented in Section II. If equation (1)
representsthe correct long run moneydemandfunction,lack of cointegrationmay be due to the fact that
we are ignoring the role of the transactionscost variable, b. Previous papers on the demand for real
balanceshave shownthat the lack of stabilitycan be due to “financialinnovation.” Where this is defined
as technological,legaland institutionalchangeswhichallowpeopleto economizeon their moneyholdings
(see Laban(1991),de Gregorioet al..(l992) and Roley(1985)). In particular,Melnick(1991)showsthat
for Israel, ignoring a proxy for financialservices leads to the misspecificationof the long run money
demandfinction and the lack of stability. I will show that the lack of stability,as is evidencedby the
lack of cointegration,of the long run moneydemandfunctionin all three of these countriesis due to the
exclusionof a proxy for financialinnovation.
If financialimovation is a relevantvariablein the demandfor moneyfinction, excludingit will
preventthe rejectionof the null hypothesisof no cointegration. Therefore, we can again have spurious
regression results in equation(7).
Previously,financialimovation has been modeledas a fall in transactionscosts representedby
an interceptdummy (Laban 1991)or a time varying intercept(Arrau and de Gregorio 1991). In this
model, I allow financialimovation to affect, a priori, the slope as well as the interceptof the money
demand function. The effects of financialimovation are capturedby a set of three dummyvariables:
i) DX which is Oprior to year X:1 and 1 thereafter
ii) DXY which is Oprior to year X:1 and equalto y thereafterand iii) DXI which is Oprior to year X:1
and equalto z thereafter. Where X is the year in each countrysamplewhere financialimovation occurs.
In BoliviaX is 1986, in Israel it is 1987, and in Venezuelait is 1989. The modifiedlong run money
demand equationbecomes:

(8)

19
The estimatedequationsfor each of the countriesare as follows:

m

mt =17.93-1 .46yl -2.93 n, 43.89D87, A.37D87Y1 +2.4D871, (9)

T=83 ~2 =.81 Dw=l.5

Israel

m, *.97+.11 y, +.04n, -8.13 D89 +.66D89Y,+.26D891, (lo)

T$5 ~2 =.82 DW=l. 1

Venezuela

m, =4.54 +1.18yt -.55 z, -22.7D86, +2.14D86Y, -.42 D861, (11)

T=47 i2=.78 DW=l.5

The standarderrors are not reportedsincethey havedegeneratedistributions. As is evident,the
long run elasticitiesof real money balanceswith respect to incomehave the right sign, but they tend to
be a-bithigh when comparedwith other estimates,exceptfor Israel. The long run elasticitiesof money
demandwith respect to incomeare 2.91 for Bolivia,0.77 for Israel, and 3.32 for Venezuela. The long
run semielasticitiesof inflationhave the right sign for both Boliviaand Venezuela, but not for Israel.
These semielasticitiesall fall between-0.5 and -1, whichis not unusualfor less developedcountries. The
long run inflation elasticities are -0.5 for BoIivia, 0.30 for IsraeI, and -0.97 for Venezuela. These
coefficientestimateswill be consistentas long as the regressors are not cointegratedseparatelyand the
dummiesare included. (See Tables 3A-3C, line 3). The R2are also very high and the DW statisticsare

20
fairly close to 2, indicatingno serious problems with serial correlation.
- of Tables3A to 3C showsthe resultsof the unit root tests on the equilibriumerrors (o’s)
of equations(9)-(11). It is clear that the dummyvariablesmatter. For all three countriesI can now reject
the nullhypothesisof no cointegration,whichindicatesthat theseequationsprovidea reasonableestimate
of the long run demandfor money in these countries. The evidencetends to support a downwardshift,
as well as a change in the slope of the long run equilibriummoney demand functionat the time when
financialinnovationoccurred in these countries. All of the DX dummiescome out negativeand large.
Having estimated a stable long run demand for money, after taking into account financial
innovation,I now estimatethe short run demandfor money and show I]owfinancialinnovationcan lead
to a faster speed of adjustmentof real balancesto changesin its determinants.

TABLE 3A: COINTEGRATION TESTS: BOLIVIA

DF(t) ADF(t) DF(nt) ADF(nt)
---------------------------------------------------------------------------------------
1. Equation(8) -2.5 -2.3 -2.1 -1.7
2. Equation(9) -5.0 -2.3 -5.1 -2.3
3.Cointegration -3.0 -1.7 -3.1 -1.9
btn. y and n
---------------------------------------------------------------------------------------
Critical Values 5% -3.5 -3.5 -2.9 -2.9
note: Four lags used for ADF

TABLE 3B: COINTEGRATION TESTS: ISRAEL

DF(t) ADF(t) DF(nt) ADF(nt)
---------------------------------------------------------------------------------------
1. Equation(8) -3.8 -2.6 -3.7 -~.6
2. Equation(9) -4.4 -3.5 -4.1 -2.8
3.Cointegration -3.4 -2.5 -1.2 -0.75
btn. y and n
---------------------------------------------------------------------------------------
Critical Values 5% -3.5 -3.5 -2.9 -2.9

The only exceptionis the regressionfor Israel that has a DW statisticof 1.1.

21
TABLE 3C: COINTEGRATION TESTS: VENEZUELA

DF(t) ADF(t) DF(nt) ADF(nt)
---------------------------------------------------------------------------------------
1. Equation(8) -1.8 -2.4 -1.8 -2.2
2. Equation(9) -5.0 -2.6 -5.0 -2.8
3. Cointegration -3.5 -3.5 -1.8 -1.2
btn. y and z
---------------------------------------------------------------------------------------
CriticalValues 5% -3.5 -3.5 -2.9 -2.9

SECTION VI: THE SHORT RUN MONEY DEMAND AND THE SPEED OF
ADJUSTMENT:AN ERROR CORRECTIONMODEL

Given the stable long run money demandwhich was estimatedin SectionV, I now turn to the

estimationof the dynamicspecificationfor real balancesin the short run. Traditionalstudiesof money

demand for other countrieshave alwaysused a partial adjustmentapproachwhich is relativelyad hoc.

Salmon(1982)demonstratesthatthis mechanismwillonlyreachthedesiredlevelof long run real balances

if this level is constantin equilibrium,which as mentionedbefore is not the case when the variablesare

nonstationary. Additionally,Stock (1987)shows that inferencesmade on estimationsby OLS when the

lagged dependentvariable is included as a regressor can lead to errors. Given these criticisms, and

followingthe RepresentationTheorem by Engle and Granger (1987), I use an error correction model

(ECM) to estimatethe short run moneydemandfor Bolivia,Israel, and Venezuela. This modelis given

by -

(12)

Where A is the first differenceoperator, ~~.lis the estimateof the “equilibriumerror” from the
...
cointegrationequations(9)-(11)and ~~is whitenoise. In order to estimateequation(12) for each of the

countriesin the sample, I follow the methodologyemployedby Hendry et al (1984). This method, of

22
goingfrom generalto specific,appliedto equation(12)impliesthatthe equationis continuouslysimplified

and reestimated. It is first estimatedin its most general form with four lags for each variableexceptthe

error correctionterm. The finalparsimoniousrepresentationis achievedafter deletingall the insignificant

variables. The results for the final version of each country’sECM are shown in Tables 4A-4C.

IITABLE 4A: AN ERROR CORRECTION MODEL FOR BOLIVIA II
II DependentVar. is Am (1) (2)
IIRegressors
constant 0.004 0.008
(0.4) (0.7)
Aq.l 0.43* 0.40*
(3.6) (3.4)
AA-3 O.So* 0.51*
(4.5) (4.8)

AYt 1.6* 1.4*
(4.8) (4.1)

AYt-3 -1.2* -1.2*
(-3.3) (-3.3)
An, -0.27* -O-3*
(-4.2) (4.42)
An,J -o.12* -o.10*
(-2.5) (-1.9)
ECtI -0.20* -0.35
(-2.6) (-3.0)
DEC,-l ------ -0.27**
(1.7)
IIR2 0.71 0.72
IISSE 0.22 0.15
T 42 42
&
t-statisticsare in parentheses, *= s]gmflcantat 5%, **= slgnlflcantat 10%

23
i

TABLE 4A: AN ERROR CORRECTION MODEL FOR ISRAEL
Dependent Var. is Am (1) (2)
Regressors
constant -0.005 -0.004
(-0.8) (-0.7)

A~.l 0.26* 0.27*
(3.9) (4.0)
Am,-~ 0.55* 0.51*
(3.2) (2.9)

AYt -0.84* 0.85*
(-9.5) (-9.6)
-0.46* -005*
AYt-3
(-4.2) (-4.3)
An, -0.29* -0.3*
(-2.9) (-3.1)
An,A -0.34* -0.33*
(-3.7) (-3.7)
EC,: -o.10* -0.09
(-2.9) (-2.7)
DEC,-l ------ -0.31**
(1.7)

R2 0.69 0.69
SSE 0.17 0.16
T 79 , 79
-. ----- *
.,.

— e, . cm && -!——:&:
-.-A.-* 1nd
t-statisticsare in parentheses, slgnlrlcan~at 3 YO, TT = SIgnlLIGdIIL
dL lU YO

...

24
TABLE 4A: AN ERROR CORRECTION MODEL FOR VENEZUELA
DependentVar. is Am (1) (2)
Regr~ors
constant -0.005 -0.006
(-0.7) (-1.0)
Amt-4 0.63* 0.57*
(6.2) (6.1)
An, -0.96* 0.66*
(-5.2) (-3.5)
EC,., -0.29* -O.16*
(-3.9) (-2.0)
DEC,l ------ -0.52*
(-3.6)
~2
0.59 0.67
SSE 0.11 0.09
T 54 54
-statlstlcsare in parentheses, *= slgmficantat 5%, **= slgnlflcantat 10%

Column (1) of each table shows the error correctionmodel for each country in its final form,

after all of the insignificantvariableshavebeen eliminated. The modelshowsencouragingresultsfor the

short run incomeand inflationelasticitiesof moneydemand. For Bolivia,the short run incomeelasticity

is 0.37, and the inflationelasticityis -0.4 which is withinthe range of previousstudiesfor less developed

countries. In addition, the coefficienton lagged money is about 0.92 which is consistentwith other

studies. For Israel, the income elasticity is 0.5 which is in line with theoretical specifications,the

inflationelasticityis -1.0, and the coefficienton lagged money is about 0.3. Finally, the results for

Venezuelaare as follows:the inflationelasticityis approximately-0.9, and the coefficienton the lagged

moneyterm is 0.63. Sinceoutputwas not a significantvariable, its elasticityis not very reliable, but it

is approximately1. All the coefficientsare highly significant.

Havingestimatedan appropriateshort run moneydemandfor these countries,I now turn to the

25
principal questionof the paper. In the model of Section II, I showedthat a decline in the transactions

costs, or what I have termed financialinnovation,causes peopleto adjusttheir real balancesfaster than

before. The error correctionmodel, ECM, of equation(12) specifiesthat the change in w dependsnot

only on the lagged values of yt and Zt, but also on the equilibriumerror that occurred in the previous

period. Viewed in this error correction framework, the matrix 6, the coefficient on o,.,, can be

interpretedas a measure of the speedby whichthe systemcorrects last period’sequilibriumerror. This

is why 5 is usuallycalledthe “adjustmentmatrix.” In Tables4A-4C, column(1) showsthat the EC term

is always of the right sign (i.e. it is negative)and highlysignificant. This impliesthat agentsare taking

into account the long run equilibrium error when adjusting their demand for real balances in the short run.

When there is a positive error (i.e. when current money holdingsare greater than the desired ones, m,

> ret”)in the long run equation,people will adjust their desired money balancesdownward.

Column(2) of Tables4A-4C show the same ECM for each countrybut includingan interactive

dummyvariable for the error correctionterm (DEC,.l). Note that in each case the coefficientincreases

in absolutevalue, and is statisticallysignificant. This confirmsempiricallywhat SectionII sets out; that

the speed of adjustmentof moneydemandto its determinantsincreaseswhenthere is financialinnovation.

In essence, people adjust faster to deviationsfrom the long run equilibriumrelation. This makes sense

since now financialinstitutionsare more efficientand people can put their money in differentassets at
.
home.

SECTION VII: COCCLUSIONS

I have shown that financialinnovationcan increasethe speed of adjustmentof money demand

to its determinants,and can lead to the instabilitytypicallyfound in other money demandstudies. This

is an aspect of the “literaturewhich has not been explored much, particularly in terms of the new

cointegrationmethods. ,.
Using the appropriatecointegrationtechniques,I have shown that by introducinga proxy for

26
financialinnovation,a stable long run money demand functioncan be obtainedfor each country. For

Bolivia,Israel, and Venezuela,the empiricalestimatesshowthatthe longrun demandfor moneynot only

shifted down, but that in the short run the effect of financialinnovationhas been to increase the speed

with which people adjust their actual money holdingsto their desired money holdings. Viewed in the

Miller-Orr transactionsdemand for money framework, a fall in the transactionscost variable means a

shorter period in which desired money balances are not equal to the optimal money balances. The

empiricalevidenceshows that the coefficienton the error correctionvariableincreasesin absolutevalue

indicatingthat once financialinnovationtakes place any disequilibriabetweendesired and actualmoney

holdingswill by eliminatedfaster.

27
REFERENCES

Arrau, Patricio, Jose De Gregorio, Carmen Reinhart and Peter Wickham, (1991), “The Demand for
Money in Developing Countries: Assessing the Role of Financial Imovation,” 1~F Workingpaper
No.91/45

Baumol,W.J (1952), “TheTransactionsDemandfor Cash:An InventoryTheoreticApproach,” Quarterly
Journal of Economics, 66, 545-556.

Ben-Bassat,A. and Marom, A. (1988), “Is the Demandfor Moneyin Israel Stable?(1965-1983),”Bank
of Israel Economic Review, N 60 (January), p.52-71.

Ben-Bassat,A. (1992), “CapitalMarketReform in Israel,” Bank of Israel Economic Review, N 65, p. 17-
30.

Blanchard,Olivier and StanleyFischer, Lectureson Macroeconomics(Cambridge,Massachusetts:MIT
Press, 1989).

Bruno,M and Meridor,L.R. (1991), “ The Costly Transitionfrom Stabilizationto SustainableGrowth:
Israel’s Case,” in Bruno M., S. Fischer, E. Helpman, N. Liviatan and L. Meridor eds.,Lessons of
EconomicStabilizationand Its Aftermath. The MIT Press

Cagan, P. (1956), “The Monetary Dynamicsof Hyperinflation,” in Studies in the Quantiw Theory of
Money, ed. M. Friedman, Chicago: Universityof ChicagoPress, pp.25-117.

Campbell,J. and Perron,P. (1991), “Pitfallsand Opportunities:What MacroeconomistsShouldKnow
About Unit Roots.” NBER Macro Annual 1991. M.I.T. Press.

Engle,R.F., andGranger,C.W.J(1987), “CointegrationandError Correction:Representation,Estimation
and Testing,” Econometric, 55, 251-276.

Engle, Robert F. and B.S. Yoo, “Forecasting and Testing Co-integrated Systems,” Journal of
Econometrics, (1987) 143-159.

Goldfeld, StephenM., The Case of Missing Money,” Brookings Papers on Economic Activi~’, (1976)
683-730.

Hafer, R. W., and Jansen, DennisW.(1991), “The Demand for Money in the United States: Evidence
from CointegrationTests,“ Journal of Money, Credit, and Banking, VOI.23, No.2 (May 1991),155-168.

Hendry, F.D., Pagan, A.R., and Sargan, J.D. (1984), “Dynamic Specification,” in Handbook of
Econometrics (VO1.2),eds. Z. Griliches and M.D. Intriligator,Amsterdam:North-Holland,pp.
1025-1100.

Laban, R. (1991) “FinancialInnovationand the Stabilityof Money Demand: Evidencefrom Argentina
and Chile,” M.I.T Chapter 3 of Ph.D. thesis

28
Leiderman,Land Marom,A. (1988), “NewEstimatesof the Demandfor Moneyin Israel,” Bank ofIsraeZ
Economic Review, N 60 (January), p.19-35.

Melnick, R.,(1988), “Two Aspects of the Demand for Money in Israel, 1970-1981,” Bank of Israel
Economic Review, N 60 (January), p.36-52.

Melnick, R., (1990), “The Demand for Money in Argentina 1978-1987:Before and After the Austral
Program,” Journal of Business and Economic Statistics, 8, N4, p427-434.

Melnick,R., (1991), “FinancialServices,Cointegration,and the Demandfor Money in Israel,” Bankof
Israel manuscript

Miller, StephenM. (1991), “MonetaryDynamics:An Applicationof Cointegrationand Error-Correction
Modeling,” Journal of Money, Credit, and Banking, VO1.23, No.2 (May 1991), 139-154.

Morales,J.A;(1991), “TheTransitionfrom Stabilizationto SustainedGrowth in Bolivia,” in Bruno M.,
S. Fischer, E. Helpman, N. Liviatan and L. Meridor eds.,Lessons of Economic Stabilizationand It$
~. The MIT Press

Morales,J.A;(1988), “La Inflacion y la Estabilizacion in Bolivia,” in Bruno M., G.Di Tella, R.
.. .
Dornbusch,and S. Fischer eds.,~nflaciony Establhzacion: La ExDerlencia de Israel. Argentina.Brasilz
~olivia v Mexico. El TrirnestreEconomicoN.62

Offenbacher,E.K. (1988), “EmpiricalStudiesof the Demandfor Money in Israel: Introduction,” Bank
of Israel Economic Review, N 60 (January), p. 1-18.

Perron, Pierre. (1988), “Trendsand RandomWalks in MacroeconomicTime Series: further Evidence
from a New Approach,” Journal of Economic Dynamics and Control 12 (1988) 297-332.

Phillips,Peter C.B.,(1986), “Understanding Spurious Regressions in Econometrics.” Journal of
Econometrics, N33,p.311-340.

Piterman, S. (1988), “TheIrreversibilityof the RelationshipBetweenInflationand Real Balances,” Bank
of Israel Economic Review, N 60 (January), p.72-83.

Stock,James H.(1987), “AsymptoticPropertiesof Least SquaresEstimatorsof Co-IntegratingVectors,”
Econometric, N 55, 1035-1056.

TheEconomistIntelligenceUnit,CountryProfile:Venezuela.Suriname.Netherlands Antilles.1988-1992.

29
International Finance Discussion Papers

IFDP
Number Titles Author(s]

1996

567 FinancialInnovationAnd The Speed of Adjustment MartinaCopelman
of MoneyDemand: EvidenceFrom Bolivia,
Israel, And Venezuela

566 Long-TermEvidenceon the Tobin and Fisher ShaghilAhmed
Effects: A New Approach - John H. Rogers

565 Some Evidenceon the Efficacy of the UK Inflation Chan Huh
TargetingRegime: An Out-of-SampleForecast
Approach

564 The Use of the Parallel MarketRate as a Guide Nita Ghei
to Settingthe Official ExchangeRate Steven B. Kamin

563 CountryFund Discountsand the MexicanCrisis of Jeffrey A. Frankel
December 1994: Did Local ResidentsTurn Sergio L. Schmukler
PessimisticBefore InternationalInvestors?

562 EasternEuropeanExport Performanceduring Nathan Sheets
the Transition SimonaBoata

561 Inflation-AdjustedPotentialOutput Jane T. Haltmaier

560 The Managementof FinancialRisksat German Allen B. Frankel
NonfinancialFirms: The Case of Metallgesellschafi David E. Palmer

559 Broad MoneyDemand and FinancialLiberalization Neil R. Ericsson
in Greece . Sunil Sharma

558 StockholdingBehaviorof U.S. Households: Evidence Carol C. Bertaut
from the 1983-89Survey of ConsumerFinances

557 Firm Size and the Impact of Profit-MarginUncertainty Vivek Ghosal
on Investment: Do FinancingConstraintsPlay a Role? Prakash Loungani

556 Regulationand the Cost of Capitalin Japan: A Case John Ammer
Study MichaelS. Gibson

555 The SovereigntyOption: The QuebecReferendumand MichaelP. Leahy
MarketVie~s on the CanadianDollar CharlesP. Thomas

Please addressrequestsfor copiesto InternationalFinanceDiscussionPapers, Divisionof
InternationalFinance, Stop 24, Board of Governorsof the FederalReserveSystem,
Washington,DC 20551.

30
InternationalFinanceDiscussionPapers

IFDP
Number Titles Author(s)

1996

554 Real ExchangeRates and Inflation in Exchange-Rate StevenB. Kamin
Based Stabilizations: An EmpiricalExamination

553 MacroeconomicState Variablesas Determinants John Ammer
of Asset Price Covariances

552 The TequilaEffect: Theory and Evidencefrom Martin Uribe
Argentina

551 The Accumulationof Human Capital: Alternative Murat F. Iyigun
Methodsand Why They Matter Ann L. Owen

550 Alternativesin Human CapitalAccumulation: Murat F. Iyigun
Implicationsfor EconomicGrowth Ann L. Owen

549 More Evidenceon the Link betweenBank MichaelS. Gibson
Health and Investmentin Japan

548 The Syndromeof Exchange-Rate-Based EnriqueG. Mendoza
Stabilizationand the UncertainDurationof Martin Uribe
CurrencyPegs

547 German Unification: What Have We Learned JosephE. Gagnon
from Multi-CountryModels? Paul R. Masson
WarwickJ. McKibbin

546 Returnsto Scale in U.S. Production: Estimates SusantoBasu
and Implications John G. Femald

545 Mexico’sBalance-of-PaymentsCrisis: A Chronicle GuillermoA. Calvo
of Death Foretold Enrique G. Mendoza

544 The Twin Crises: The Causesof Bankingand GracielaL. Kaminsky
Balance-of-PaymentsProblems Carmen M. Reinhti

543 High Real Interest Rates in the Aftermathof GracielaL. Kaminsky
Disinflation: Is it a Lack of Credibility? LeonardoLeiderrnan

542 PrecautionaryPortfolio Behaviorfrom a Life-Cycle Carol C. Bertaut
Perspective MichaelHaliassos

541 Using OptionsPrices to Infer PDF’s for Asset Prices: WilliamR. Melick
An Applicationto Oil Prices During the Gulf Crisis CharlesP. Thomas

31
International Finance DiscussionPapers

IFDP
Number Titles Author(s)

1996

540 MonetaryPolicy in the End-Gameto Exchange-Rate Steven B. Kamin
Based Stabilizations: The Case of Mexico John H. Rogers

539 Comparingthe WelfareCosts and the Initial Dynamics Martin Uribe
of AlternativeTemporaryStabilizationPolicies

538 Long Memory in Inflation Expectations: Evidence Joseph E. Gagnon
from InternationalFinancialMarkets

537 Using Measuresof Expectationsto Identi$ the Allan D. Brunner
Effects of a MonetaryPolicy Shock

536 Regime Switchingin the DynamicRelationship Chan Huh
betweenthe Federal Funds Rate and Innovationsin
NonborrowedReserves

535 The Risks and Implicationsof ExternalFinancial Edwin M. Truman
Shocks: Lessonsfrom Mexico

534 CurrencyCrashesin EmergingMarkets: An Jefiey A. Frankel
EmpiricalTreatment Andrew K. Rose

533 RegionalPatterns in the Law of One Price: The CharlesEngel
Roles of GeographyVs. Currencies John H. Rogers

32